From dc2e8928c18db3004700794e50a31ac4504b4a31 Mon Sep 17 00:00:00 2001
From: Taylor <quantumtraveling@gmail.com>
Date: Wed, 11 Jun 2025 04:19:38 -0400
Subject: [PATCH] Pass --keep to llama-server

---
 tools/server/server.cpp | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/tools/server/server.cpp b/tools/server/server.cpp
index 3b5e03528e2d7..1b1cf439baa5f 100644
--- a/tools/server/server.cpp
+++ b/tools/server/server.cpp
@@ -233,6 +233,7 @@ struct server_task {
         slot_params defaults;
         defaults.sampling    = params_base.sampling;
         defaults.speculative = params_base.speculative;
+        defaults.n_keep      = params_base.n_keep;
 
         // enabling this will output extra debug information in the HTTP responses from the server
         params.verbose           = params_base.verbosity > 9;
@@ -2060,6 +2061,7 @@ struct server_context {
             SLT_INF(slot, "new slot n_ctx_slot = %d\n", slot.n_ctx);
 
             slot.params.sampling = params_base.sampling;
+            slot.params.n_keep = params_base.n_keep;
 
             slot.callback_on_release = [this](int) {
                 queue_tasks.pop_deferred_task();